CN104636447A - Intelligent evaluation method and system for medical instrument B2B website users - Google Patents

Abstract

The invention belongs to the field of information technology, and particularly discloses an intelligent evaluation method and system for medical instrument B2B website users. The evaluation method comprises the steps of 1, user evaluation index base building; 2, evaluation rule base modeling, 3, evaluation executing and 4, output and feedback. The evaluation system comprises four corresponding modules which execute the four steps respectively. The basic information, historical transaction information and online comment information of the medical instrument B2B website users are synthetically computed and modeled, a quantized evaluation value is provided for each site user and used as a reference during online transaction, and therefore the purposes of evaluating the transaction concluding possibility and preventing potential risks are achieved. The intelligent evaluation method and system for the medical instrument B2B website users have the advantages that data mining, multiple regression modeling and natural language processing technologies are comprehensively utilized for site user modeling evaluation objectively and comprehensively; a feedback mechanism is provided, an intelligent evaluation rule base can be continuously improved, and therefore evaluation results are more accurate.

Description

A kind of Intelligent Evaluation method and system towards medicine equipment B2B websites user

Technical field

The invention belongs to areas of information technology, be specifically related to website user's assessment technique, particularly a kind of Intelligent Evaluation method and system towards medicine equipment B2B websites user.

Background technology

In recent years, along with the development of the communication technology and the remarkable growth of Internet Users, the various application based on internet are arisen at the historic moment, and bring great convenience to daily life.The advantages such as the cross-region that shopping online possesses because of it, interaction and round-the-clock property are subject to the favor of more and more consumer.The sales mode comparing solid shop/brick and mortar store due to merchandising on the internet has some characteristics and advantage, and thus cause many goods providers and touch net one after another, medical supplies business is no exception.In China, medicine equipment is as more special commodity, it carries out often through the medicine equipment B2B websites possessing " internet medicine transactional services qualification certificate " in online spending, medicine equipment both parties carry out the online transaction of medicine equipment by this channel, intermediate links can be greatly reduced, cost-saving and reduction buying difficulty.But, along with the growth of medicine equipment B2B websites registered user number, and the increasing of online transaction number of times, user is sincere to be highlighted with transaction risk problem.In order to assess the conclusion of the business possibility of online transaction better and prevent transaction risk as much as possible, science is carried out to website registered user, objective appraisal is an extremely important and significant job.

At present, the essential information of filling in when registering mainly with user greatly the evaluation of medicine equipment B2B websites user is as main, or the credit appraisal of indiscriminately imitating in some online seller's performance assessment criteria, this evaluation is often made a report on by the opposing party of every transaction, by the result that this evaluation method is given, to a certain extent the risk of transaction is estimated and there is meaning, but, also there is certain limitation in above-mentioned evaluation method, owing to only considering user's registration information or seller's credit, therefore, the following problem of ubiquity:

(1) evaluation index is comparatively single, cannot comprehensive, objective reaction user real conditions, is main to the evaluation degree of adoption adjective of website user, unclear boundaries between grade, and inconvenience is intuitively understood;

(2) evaluation rule storehouse used lacks feedback and automatic update mechanism, easily occurs evaluating hysteresis phenomenon, cause obtain and evaluate with user's truth gap large, accuracy is not high.

Summary of the invention

The object of the invention is the deficiency evaluated for existing medicine equipment B2B websites user, comprehensive utilization data mining, multiple regression modeling and natural language processing technique calculate and modeling the essential information of website user, historical transactional information, line Evaluation opinion, to the evaluation of estimate that each user one of website quantizes, to reach objective, comprehensively and visual evaluation user, be prevention online transaction risk object for referencial use.

For achieving the above object, the present invention adopts following technical scheme, comprises following content:

1, towards a medicine equipment B2B websites user's Intelligent Evaluation method and system, it is characterized in that, evaluation method comprises 4 steps, is specially:

Step 1: set up user's evaluation index storehouse;

Comprise successively: B2B websites user data excavates, user's evaluation index is chosen and formed totally 3 processes with user's index storehouse, wherein:

B2B websites user data excavates: carry out the operation of data prediction, natural language processing, classification by the essential information to medicine equipment B2B websites user, historical transactional information, on the net review information, and the flow process of the K means clustering algorithm of historical transactional information employing improvement being carried out to cluster is as follows:

Flow process 1: data prediction; By transaction (T every in the historical transactional information of website _i) in product total quantity (Q), transaction total charge (M), complete the number of days (D) of payment, the scoring (S) of the rear the other side that concluded the business, enter row systemcount and export in Excel;

Flow process 2: initialization cluster centre; The cluster number of clusters that setting will divide is k, and bunch center C of this k cluster _j(Q, M, D, S), j=1,2 ..., k;

Flow process 3: start circulation, calculates Euclidean distance and classifies; Enter circulation, calculate every transaction (T _i) in product total quantity (Q), transaction total charge (M), complete the number of days (D) of payment, the scoring (S) of the rear the other side that concluded the business to the Euclidean distance D (T at k clustering cluster center _i, C _j), i=1,2 ..., n, j=1,2 ..., k; If meet D (T _i, C _k)=min{D (T _i, C _j), j=1,2 ..., n}, be then divided into the most close class bunch;

Flow process 4: recalculate the average of every class in order to determine new cluster centre; The computing formula of new cluster centre is as follows:

$C_j^{k+1}(Q,M,D,S)=\frac{1}{n_j}\underset{i=1}{\overset{n_j}{\mathrm{\Σ}}}{T_i}^{\left(j\right)}$

In formula, cluster centre, n _jthe number of samples comprised in a jth Clustering Domain n, T _i ^(j)it is the every transaction in a jth clustering cluster;

Flow process 5: calculate the error sum of squares of every class and judge; The computing formula of error sum of squares is as follows:

$J=\underset{j=1}{\overset{k}{\mathrm{\Σ}}}\underset{k=1}{\overset{n_j}{\mathrm{\Σ}}}{\left|\right|T_k^j-C_j^k(Q,M,D,S)\left|\right|}^2$

In formula, J is error sum of squares criterion function, n _jthe number of samples comprised in a jth Clustering Domain n, T _i ^(j)the k transaction in a jth clustering cluster, it is the cluster centre of a jth clustering cluster;

Judge whether J restrains, if convergence, then terminate and jump out circulation; Otherwise circulation adds 1, Returning process 3, continue to calculate the individual new cluster centre of k;

Flow process 6: k the cluster set exporting historical transactional information;

User's evaluation index is chosen: take Delphi expert opinion method, determine that index that user evaluates is marked by registered capital, registration time length, transaction count, dealing money, transaction according to the suggestion of expert feedback, service scoring, sincere scoring, message scoring, be punished number of times totally 9 indexs form;

User's evaluation index storehouse is formed: to selected 9 user's evaluation indexes, adopts AHP analytical hierarchy process agriculture products weight, forms medicine equipment B2B websites user's index storehouse;

Step 2: evaluation rule storehouse modeling;

Comprise successively: multiple regression modeling, artificial intelligence modeling and Intelligent Evaluation rule base form totally three processes, wherein:

Multiple regression modeling: quantize medicine equipment B2B websites user comprehensive evaluation by multivariate regression model and divide, its multivariate regression model form is as follows:

Y＝α+β _i*X _i

In formula, Y refers to that comprehensive evaluation is divided, and α is intercept item, and i gets 1 to 9, i.e. 9 indexs; β refers to regression coefficient, is drawn by least square method estimation; X _irefer to regression variable, that is: can for returning 9 index values calculated after data prediction;

Artificial intelligence modeling: use BP neural network to carry out sample training and modeling to the desired value in website user's index storehouse, comprising: the planned network number of plies, design neuromere are counted, Designing Transfer Function and learning function totally 3 flow processs, specific as follows:

Flow process 1: the planned network number of plies; Consider that 3 layers of BP neural network can approach any mapping relations with arbitrary accuracy, therefore, the number of plies of BP neural network is chosen to be 3 layers, that is: input layer, hidden layer and output layer;

Flow process 2: design neuromere is counted; Input layer number is set to 9, that is: 9 indexs; Output layer nodes is set to 1, that is: the comprehensive evaluation that output multiple regression modeling obtains is divided; Node in hidden layer is by experimental formula and repetition training provides, in formula, I is input layer number, and O is output layer nodes, and n is made up of the integer of 1 to 10;

The square error computing formula of neural network is as follows:

$\mathrm{MSE}=\frac{1}{\mathrm{ns}}\underset{s=1}{\overset{s}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{(\widehat{y_{\mathrm{sj}}}-y_{\mathrm{sj}})}^2$

In formula, MSE is the square error of whole BP neural network, and n is output node sum, and s is the sum of training sample, the desired output of BP neural network, y _sjit is the real output value of BP neural network;

Flow process 3: Designing Transfer Function and learning function; Select tansig as hidden layer neuron transport function; Select purelin as output layer neural transferring function; Select traingdx as training function; Adopt 0.1 as learning rate initial value; Employing 0.9 is as the initial value of factor of momentum;

Intelligent Evaluation rule base is formed: on the basis of multiple regression modeling and artificial intelligence modeling, extracting rule sets up Intelligent Evaluation rule base table " tb_AssessRule " for Intelligent Evaluation in database " Database_B2B_MIA ", the list structure of " tb_AssessRule " comprises number of regulation, content, rule type, confidence level totally four fields, wherein:

Number of regulation: database design becomes the mode automatically increasing 1, and initial value is 1;

Content: represent by varchar (200) type, changes into rule format by the rule of modeling gained before and is saved in database;

Rule type: represent by varchar (4) type, and represent that negative sense is correlated with 0,1 represents that forward is correlated with;

Confidence level: adopt numeric (8,4) type to represent, record the credibility number percent of every rule;

Step 3: perform evaluation;

Comprise successively: choose user and carry out evaluating totally two processes, wherein:

Choose user: choose object to be evaluated from medicine equipment B2B websites, namely complete the user of website registration;

Evaluate: call the rule in Intelligent Evaluation rule base, carry out the automatic Evaluation coupling of similarity;

Step 4: export and feed back;

Comprise successively: Output rusults and renewal Intelligent Evaluation rule base totally two processes, wherein:

Output rusults: export the score value that selected user provides after Intelligent Evaluation, score range is 0 to 100;

Upgrade Intelligent Evaluation rule base: by this evaluation rule result feedback to Intelligent Evaluation rule base, and use trigger automatically to upgrade corresponding rule.

2, based on the intelligent evaluation system towards medicine equipment B2B websites user according to claim 1, it is characterized in that comprising four modules: set up user's evaluation index library module, evaluation rule storehouse MBM, perform evaluation module, export and feedback module, these four modules perform respectively and correspond to towards 4 steps in the Intelligent Evaluation method of medicine equipment B2B websites user, wherein:

Describedly set up user's evaluation index library module, comprise that B2B websites user data excavates, user's evaluation index is chosen and formed totally 3 submodules with user's index storehouse, these 3 submodules are held respectively row powerprofit requires the function of 3 processes in 1 step 1;

Described evaluation rule storehouse MBM, comprise multiple regression modeling, artificial intelligence modeling and Intelligent Evaluation rule base and form totally 3 submodules, these 3 submodules are held respectively row powerprofit requires the function of 3 processes in 1 step 2;

Described execution evaluation module, comprise and choose user and carry out evaluating totally 2 submodules, these 2 submodules are held respectively row powerprofit requires the function of 2 processes in 1 step 3;

Described output feedback module, comprise Output rusults and upgrade Intelligent Evaluation rule base totally 2 submodules, these 2 submodules are held respectively row powerprofit requires the function of 2 processes in 1 step 4.

Accompanying drawing explanation

fig. 1it is overall architecture of the present invention figure.

fig. 2it is participle interface under R visual edit instrument RStuio of the present invention figure.

fig. 3that Intelligent Evaluation rule base of the present invention forms signal figure.

fig. 4it is the flow process of the invention process cluster data mining method figure.

Embodiment

Referring to accompanying drawing, various enforcement of the present invention is described in further detail.

fig. 1show overall architecture of the present invention figure.The present invention is made up of four steps, namely sets up user's evaluation index storehouse (1), evaluation rule storehouse modeling (2), performs evaluation (3) and export and feed back (4).Wherein, it is adopt Delphi Specialist Research method that the user's evaluation index storehouse set up in user's evaluation index storehouse (1) step forms (5), select registered capital, registration time length, transaction count, dealing money, transaction scoring, service scoring, sincere scoring, message scoring, be punished number of times totally 9 indexs to evaluate medicine equipment B2B websites user, and adopt AHP analytical hierarchy process, agriculture products weight, comprises following 5 processes:

Process 1: set up analysis level structural model; Wherein ground floor is divided into log-on message, Transaction Information, information on services, rewards and punishments information four class; The second layer is 9 indexs, specifically as table 1shown in.

table 1

Process 2: development of judgment matrix;

Adopt Consistent Matrix method, index carried out mutual comparing between two, constructs judgment matrix M as follows:

$M=\left[\begin{array}{cccc}1& 1/2& \·\·\·& 1/4\\ 1/4& 1& \·\·\·& 1/6\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ 3& 1& \·\·\·& 2\end{array}\right]$

Process 3: calculate each element weights of judgment matrix;

Each row of judgment matrix M is normalized calculating, wherein, M _ithe geometrical mean of this row element, by formula calculate, gained column vector A=(A ₁, A ₂..., A _n) ^tas the weight vectors of judgment matrix, i.e. each judging quota A ₁₁, A ₁₂, B ₁₁...., D ₁₁weight.

Process 4: to the consistency check of judgment matrix;

The judgment matrix coincident indicator CI obtained checks, and its computing formula is: in formula, λ _maxthe eigenvalue of maximum of judgment matrix, then, table look-up to obtain corresponding random index RI, then pass through formula calculate consistency ration (CR); If CR<0.1, the consistance of judgment matrix is acceptable, otherwise judgment matrix does not meet coherence request, need again revise.This calculates CR=0.00068, meets CR<0.1, and the consistance of judgment matrix M is acceptable.

Process 5: provide weighted list;

Finally, after above-mentioned calculating, the weight allocation of each index as table 2shown in.

Index	A 11	A 12	B 11	B 12	B 13	C 11	C 12	C 13	D 11
										Weight	0.095	0.173	0.091	0.088	0.123	0.074	0.071	0.132	0.153

B2B websites user data excavates (6) and refers to that maintenance data digging tool is to user basic information, historical transactional information and online review information are excavated, with online review information, " delivery is rapid, product is certified products " carry out excavation for example, first, undertaken by natural language processing instrument R language, participle bag is called by code " library (Rwordseg) " in R language visual handling implement RStudio, and with code, " (' delivery is rapid for segmentCN, product is certified products ') " perform participle, this line Evaluation opinion is just divided into 5 parts automatically like this, i.e.: " rapid product of delivering is certified products ", then each participle will completed, does polarities match with " positive-negative polarity lexicon " respectively and calculates, and add up positive-negative polarity frequency, when front polarity statistical value is greater than negative polarity statistical value, is 1 in a database to this comment assignment, otherwise, when front polarity statistical value is less than negative polarity statistical value, be 0 to this comment assignment in a database, in this example, front polarity statistical value is greater than negative polarity statistical value, and therefore online review information " delivery rapidly, product be certified products " assignment is 1.

It is complete on the basis of multiple regression modeling (9) and artificial intelligence modeling (10) that Intelligent Evaluation rule base forms (8), by setting up Intelligent Evaluation rule base table " tb_AssessRule " for Intelligent Evaluation in " Database_B2B_MIA " database, the list structure of " tb_AssessRule " comprises number of regulation, content, rule type, confidence level is totally four fields, such as: typically rule: the confidence level that { " hour of log-on <1 " and " transaction scoring be 5 points " and " after sale service scoring is 5 points " }->{ transaction exists risk is 76%}.

fig. 2show participle interface under R visual edit instrument RStudio of the present invention figure.Under RStudio environment, by menu File->Open File, import the code file " mytest.R " of the online comment text of process, this code file can carry out participle, polarity judgement process to 520 line Evaluation opinions in " 47.txt " text." mytest.R " is presented in the workspace of the upper left of RStudio environmental interface after opening, this interface upper right portion display history exports, bottom left section is that worktable exports, lower right-most portion shows the bag installed, " mytest.R " code file is performed at this interface, article 520, line Evaluation opinion will be performed participle by function RWordseg (), until after function TotalPolar () statistics positive-negative polarity word frequency, export the result of 0 or 1.

fig. 3show Intelligent Evaluation rule base of the present invention and form signal figure.In the user basic information " Database_B2B_MIA " database, historical transactional information and online review information, the information structure table " tb_AssessRule_Processing " of user ID, registered capital, registration time length, transaction count, dealing money, transaction scoring, service scoring, sincere scoring, message scoring, number of times of being punished is extracted by call function ExtractInformatin () in table" confidence level " provides initial value by expert survey, and the logical code of function ExtractInformatin () realizes seeing annex 1.Then maintenance data digging technology carries out rule digging to " tb_AssessRule_Processing ", and once search out rule, to be just saved in " Intelligent Evaluation rule base " inner.The support vector machine method his-and-hers watches " tb_AssessRule_Processing " in data mining are such as adopted to classify, the code sample adopting support vector machine method in data mining to realize prediction in matlab is shown in annex 2, in this example, support vector machine selected parameter is C=1000, ε=0.01, σ=1, its kernel function K (x _i, x) adopt gaussian radial basis function kernel function, computing formula is as follows:

$K(x_i,x)=\exp (-\frac{\left|\right|x_i-x\left|\right|}{{2\mathrm{\&sigma;}}^2}$

fig. 4show the flow process of the invention process cluster data mining method figure.Wherein:

As shown in flow process 15, initialization cluster number k and iterations n, such as: poly-10 classes, iterations is 500;

As shown in flow process 16, treat clustering object (user's evaluation information data object) and carry out k cluster, and calculate k cluster centre, such as: each user's evaluation information data object comprises registered capital, registration time length, transaction count, dealing money, transaction scoring, service scoring, sincere scoring, message scoring, number of times totally 9 attributes of being punished, the center of 10 classes is polymerized to for { C _1i, C _2i, C _3i, C _4i, C _5i, C _6i, C _7i, C _8i, C _9i, i=1,2 ..., 10;

As shown in flow process 17, all objects are sorted out, according to nearby principle, such as: 290 users are referred in 10 classes;

As shown in flow process 18, recalculate the cluster centre in new classification, after namely 290 users are referred to 10 classes, mean value computation are carried out to the every class of 10 class of newly returning, obtains 10 new cluster centres ₁;

As shown in flow process 19, judge whether object restrains, namely judge whether ε is less than the threshold value of setting, namely represents convergence if be less than, enter flow process 20; Otherwise Returning process 17, sorts out all objects again;

As shown in flow process 20, export the result of this cluster;

As shown in flow process 21, judge whether to reach maximum iteration time, if reach maximum iteration time, enter flow process 22, otherwise Returning process 17, again all objects are sorted out, such as: whether reach 500 times set iteration, if do not reach 500 times, then by iterations cumulative 1, turn back to flow process 17;

As shown in flow process 22, computing terminates to return cluster result.

Annex

Annex 1

ExtractInformatin () the function logics code extracting evaluation index information is as follows:

Annex 2

The code sample adopting support vector machine method in data mining to realize predicting in matlab is as follows:

。

Claims (2)

1. towards a medicine equipment B2B websites user's Intelligent Evaluation method and system, it is characterized in that, evaluation method comprises four steps, is specially:

Step 1: set up user's evaluation index storehouse;

Comprise successively: B2B websites user data excavates, user's evaluation index is chosen and formed totally three processes with user's index storehouse, wherein:

B2B websites user data excavates: carry out the operation of data prediction, classification by the essential information to medicine equipment B2B websites user, historical transactional information, on the net review information, the flow process adopting the K means clustering algorithm improved to carry out cluster to historical transactional information is as follows:

Flow process 1: data prediction; By transaction (T every in the historical transactional information of website _i) in product total quantity (Q), transaction total charge (M), complete the number of days (D) of payment, the scoring (S) of the rear the other side that concluded the business, carry out adding up and export in Excel;

Flow process 2: initialization cluster centre; The cluster number of clusters that setting will divide is k, and bunch center C of this k cluster _j(Q, M, D, S), j=1,2 ..., k;

Flow process 3: start circulation, calculates Euclidean distance and classifies; Enter circulation, calculate every transaction (T _i) in product total quantity (Q), transaction total charge (M), complete the number of days (D) of payment, the scoring (S) of the rear the other side that concluded the business to the Euclidean distance D (T at k clustering cluster center _i, C _j), i=1,2 ..., n, j=1,2 ..., k; If meet D (T _i, C _k)=min{D (T _i, C _j), j=1,2 ..., n}, be then divided into the most close class bunch;

Flow process 4: recalculate the average of every class in order to determine new cluster centre; The computing formula of new cluster centre is as follows:

$C_j^{k+1}(Q,M,D,S)=\frac{1}{n_j}\underset{i=1}{\overset{n_j}{\mathrm{\&Sigma;}}}{T_i}^{\left(j\right)}$

In formula, cluster centre, n _jthe number of samples comprised in a jth Clustering Domain n, T _i ^(j)it is the every transaction in a jth clustering cluster;

Flow process 5: calculate the error sum of squares of every class and judge; The computing formula of error sum of squares is as follows:

$J=\underset{j=1}{\overset{k}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{n_j}{\mathrm{\&Sigma;}}}{\left|\right|T_k^j-C_j^k(Q,M,D,S)\left|\right|}^2$

In formula, J is error sum of squares criterion function, n _jthe number of samples comprised in a jth Clustering Domain n, T _i ^(j)the k transaction in a jth clustering cluster, it is the cluster centre of a jth clustering cluster;

Judge whether J restrains, if convergence, then terminate and jump out circulation; Otherwise circulation adds 1, Returning process 3, continue to calculate the individual new cluster centre of k;

Flow process 6: k the cluster set exporting historical transactional information;

User's evaluation index is chosen: take Delphi expert opinion method, determine that index that user evaluates is marked by registered capital, registration time length, transaction count, dealing money, transaction according to the suggestion of expert feedback, service scoring, sincere scoring, message scoring, be punished number of times totally 9 indexs form;

User's evaluation index storehouse is formed: to selected 9 user's evaluation indexes, adopts AHP analytical hierarchy process agriculture products weight, forms medicine equipment B2B websites user's index storehouse;

Step 2: evaluation rule storehouse modeling;

Comprise successively: multiple regression modeling, artificial intelligence modeling and Intelligent Evaluation rule base form totally three processes, wherein:

Multiple regression modeling: quantize medicine equipment B2B websites user comprehensive evaluation by multivariate regression model and divide, its multivariate regression model form is as follows:

Y＝α+β _i*X _i

In formula, Y refers to that comprehensive evaluation is divided, and α is intercept item, and i gets 1 to 9, i.e. 9 indexs; β refers to regression coefficient, is drawn by least square method estimation; X _irefer to regression variable, that is: can for returning 9 index values calculated after data prediction;

Artificial intelligence modeling: use BP neural network to carry out sample training and modeling to the desired value in medicine equipment B2B websites user's index storehouse, comprise: the planned network number of plies, design neuromere are counted, Designing Transfer Function and learning function totally 3 flow processs, specific as follows:

Flow process 1: the planned network number of plies; Consider that 3 layers of BP neural network can approach any mapping relations with arbitrary accuracy, therefore, the number of plies of BP neural network is chosen to be 3 layers, that is: input layer, hidden layer and output layer;

Flow process 2: design neuromere is counted; Input layer number is set to 9, that is: 9 indexs; Output layer nodes is set to 1, that is: the comprehensive evaluation that output multiple regression modeling obtains is divided; Node in hidden layer is by experimental formula and repetition training provides, in formula, I is input layer number, and O is output layer nodes, and n is made up of the integer of 1 to 10;

The square error computing formula of neural network is as follows:

$\mathrm{MSE}=\frac{1}{\mathrm{ns}}\underset{s=1}{\overset{s}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{({\hat{y}}_{\mathrm{sj}}-y_{\mathrm{sj}})}^2$

In formula, MSE is the square error of whole BP neural network, and n is output node sum, and s is the sum of training sample, the desired output of BP neural network, y _sjit is the real output value of BP neural network;

Flow process 3: Designing Transfer Function and learning function; Select tansig as hidden layer neuron transport function; Select purelin as output layer neural transferring function; Select traingdx as training function; Adopt 0.1 as learning rate initial value; Employing 0.9 is as the initial value of factor of momentum;

Intelligent Evaluation rule base is formed: on the basis of multiple regression modeling and artificial intelligence modeling, extracting rule sets up Intelligent Evaluation rule base table " tb_AssessRule " for Intelligent Evaluation in database " Database_B2B_MIA ", the list structure of " tb_AssessRule " comprises number of regulation, content, rule type, confidence level totally four fields, wherein:

Number of regulation: database design becomes the mode automatically increasing 1, and initial value is 1;

Content: represent by varchar (200) type, changes into rule format by the rule of modeling gained before and is saved in database;

Rule type: represent by varchar (4) type, and represent that negative sense is correlated with 0,1 represents that forward is correlated with;

Confidence level: adopt numeric (8,4) type to represent, record the credibility number percent of every rule;

Step 3: perform evaluation;

Comprise successively: choose user and carry out evaluating totally two processes, wherein:

Choose user: choose object to be evaluated from medicine equipment B2B websites, namely complete the user of website registration;

Evaluate: call the rule in Intelligent Evaluation rule base, carry out the automatic Evaluation coupling of similarity;

Step 4: export and feed back;

Comprise successively: Output rusults and renewal Intelligent Evaluation rule base totally two processes, wherein:

Output rusults: export the score value that selected user provides after Intelligent Evaluation, score range is 0 to 100;

Upgrade Intelligent Evaluation rule base: acquired results is fed back to Intelligent Evaluation rule base, and use trigger automatically to upgrade corresponding rule.

2. based on the intelligent evaluation system towards medicine equipment B2B websites user according to claim 1, it is characterized in that comprising four modules: set up user's evaluation index library module, evaluation rule storehouse MBM, perform evaluation module, export and feedback module, these 4 modules perform respectively and correspond to towards four steps in the Intelligent Evaluation method of medicine equipment B2B websites user; Wherein:

Describedly set up user's evaluation index library module, comprise that B2B websites user data excavates, user's evaluation index is chosen and formed totally 3 submodules with user's index storehouse, these 3 submodules respectively enforcement of rights require the function of 3 processes in 1 step 1;

Described evaluation rule storehouse MBM, comprise multiple regression modeling, artificial intelligence modeling and Intelligent Evaluation rule base and form totally 3 submodules, these 3 submodules respectively enforcement of rights require the function of 3 processes in 1 step 2;

Described execution evaluation module, comprise and choose user and carry out evaluating totally 2 submodules, these 2 submodules respectively enforcement of rights require the function of 2 processes in 1 step 3;

Described output feedback module, comprise Output rusults and upgrade Intelligent Evaluation rule base totally 2 submodules, these 2 submodules respectively enforcement of rights require the function of 2 processes in 1 step 4.